Three candidates, one weighted verdict.
How dementia, ADHD and anxiety were scored against six weighted criteria until one option won on the evidence ā plus the personas, problem statement and How-Might-We questions that convergence produced, and the one finding that complicates the win.
Source: DESIGN_THINKING_FRAMEWORK.md Ā§2 DEFINE (2.1 Personas Ā· 2.2 JTBD Ā· 2.3 scoring matrix Ā· 2.4 POV/HMW) Ā· interviews/user-interviews.md (5 reconstructed semi-structured interviews + cross-interview synthesis)
Six criteria, weighted before any option was scored
Convergence didn't start with a favourite. Five-plus candidate directions had entered the funnel from Empathize and brainstorming ā dementia, ADHD, anxiety, plus sleep, chronic pain and depression noted as further extras ā and the Define phase's job was to score them against fixed criteria, not to pick a winner by fiat. Three territories carried enough evidence to be scored in full: dementia, ADHD and anxiety.
Each criterion was weighted in advance, reflecting what actually determines whether a corporate-asset play is fundable ā clinical grounding, reachable demand, monetization, fit with what the company already owns, regulatory exposure, and synergy with the wearable hardware already in hand.
Weights sum to 100%. "Regulatory ease" is scored in the source matrix as regulatory risk ā a low-risk option scores well here.
The full scoring matrix
Qualitative scores per criterion, per territory, as entered in the Define-phase matrix.
| Criterion (weight) | Dementia | ADHD | Anxiety |
|---|---|---|---|
| Clinical evidence strength for music listening (25%) | Medium ā but only for FAMILIAR music (Dr. Wang + literature), weak fit for AI-generated new music | Weak/mixed | Strong (meta-analyses, HRV/cortisol studies) |
| Reachable market & self-serve adoption (20%) | Low ā user ā buyer ā payer; institutional sales | Medium | High ā self-aware, self-paying, huge prevalence |
| Willingness to pay / monetization (20%) | Low in China (Dr. Wang) | Medium | Medium-high, esp. US/EU subscription norms |
| Fit with company assets: AI music + catalog + distribution (15%) | Poor (needs THEIR old songs ā licensing hell) | Medium | Strong (adaptive versions of real catalog music) |
| Regulatory risk at wellness positioning (10%) | Medium | High (ADHD = medical claim, EndeavorRx precedent = FDA land) | Low if positioned as stress-relief wellness |
| Wearable / AI-hardware synergy (10%) | Low | Medium | High (HRV is literally a stress metric) |
Weighted totals
The same matrix, under uncertainty ā and after the interview evidence
Point scores hide how sure we are. In v2 every cell is a low / mode / high range, the weights are jittered Ā±40% and 5,000 Monte-Carlo draws report how often each option ranks first. Anxiety's asset fit (4 ā 3) and wearable synergy (5 ā 4) were revised down after the acute-state interviews and the HRV-API facts: mode total 4.30 (was 4.55), still first in > 99% of draws. The choice is robust; what changed is which anxiety product.
Open the interactive Monte-Carlo and weight tornado ā Ā· and the belief register (A1 split by arousal state) ā
Who convergence is for, and the job they're hiring us for
Hybrid-work professional
28ā40, mild-to-moderate anxiety, owns a smartwatch, already self-medicates with music/podcasts, won't see a therapist.
Sleep-anxious new parent
The framework's second segment ā carried forward as an example alongside the primary persona for future validation.
Diagnosed severe GAD patient
Belongs to clinical care, not us ā named explicitly so the product doesn't drift into a claim it can't support.
Jobs-to-be-Done
"When I feel my chest tighten during the workday at home, I want something that calms me within minutes without requiring effort or looking like 'therapy', so I can get back to work."
Point-of-View statement
How Might We ā the questions convergence produced
Five HMWs carried forward from the POV into Ideate.
The win comes with a complication
HMW #1 assumes users want "real music, not soundscapes." The first three anxiety interviews said the opposite.
Our unfair advantage points one way; the acute user need points the other.
All three acute-anxiety interviewees ā Zhang Wei (I-03, 28, backend engineer), Lin Jing (I-04, 26, new-media exec) and Mr. Chen (I-05, 35, startup founder) ā independently rejected melodic, lyrical, rhythmic music when anxious and asked for the opposite: featureless, non-melodic, lyric-free, emotionally-neutral, low-information sound. None asked for songs they love; two explicitly said melody/lyrics/rhythm made their anxiety worse.
"äøē§å®å Ø'ē©ŗę“'ćę²”ęę ē»Ŗčµ·ä¼ćåäøå µęč½Æēå¢äøę ·ē声é³,ęęč·äøēéē¦»å¼" ā a completely empty, emotionless, "soft wall" of sound that isolates me from the world. ā I-03, Zhang Wei
"č½č®©ęčēµę³¢å¹³ē¼äøę„ē声é³,äøč¦ęęå¾,äøč¦č®©ęäŗ§ēä»»ä½ē»é¢ęåčę³" ā flatten my brainwaves down; no melody, no imagery or association. ā I-04, Lin Jing
"ę¢č½å”«č”„ē©ŗé“ē©ŗē½,åč½åø¦ę„ē»åƹ平éē声é³" ā a sound that fills the emptiness AND brings absolute calm. ā I-05, Mr. Chen
Neutral, non-melodic, adaptive sound ā Endel's territory. A "soft wall," "flattened brainwaves," "fill the emptiness." Does not lever the catalog.
Real licensed music + artists + AI production pipeline ā the company's unfair advantage, built to make "calm versions of real songs," which is exactly what acute users say they don't want.
Existing neutral audio has its own complaints the interviews surfaced: white noise/nature has sharp jarring frequencies (bird calls, high piano) [I-03]; brainwave/Alpha tracks with a low hum caused nausea [I-05]; "healing" playlists feel fake and trigger reactance [I-04]; anything with a beat makes the heart entrain to the rhythm ā physiologically counterproductive [I-04].
Resolution to test ā segment by arousal state: featureless adaptive sound for acute/panic/sleep-onset; preferred/familiar real music only for low-arousal daytime wind-down or mood repair, where the music-anxiety literature and Dr. Wang's familiarity insight still apply. This reading is corroborated (and tempered) by the n=200 netnography referenced in the same source: both preferences coexist across App Store/Product Hunt reviews of Calm, Endel and Brain.fm ā some fault Endel/Brain.fm for not being "the song I expected," others value the absence of lyrics precisely because there are "no hooks." The correct read is two distinct jobs/segments, not a single content bet.
Caveat: n=3 for anxiety, possibly self-selected, describing acute states only ā a milder "unwind after work" need may still favour real music. Treat as hypothesis-generating. The biometric-loop hypothesis (adapt to the watch, show the measured result) is not challenged by this finding ā only the content choice (real songs vs neutral sound) is, and that is what the nā„20 efficacy pilot and concept test must now A/B directly.